The effect of knowledge spillovers on regional new firm formation: The Greek manufacturing case

Abstract

This research examines the effect of knowledge spillovers on new firm formation across Greek regions in manufacturing over the period 2002–2010. The econometric analysis results reveal that knowledge spillovers, as proxied by innovation and high-tech labor measures, positively affect regional new firm formation rates. Intra-sectoral spillovers, as captured by geographic sectoral specialization and industrial intensity, also positively affect regional new firm formation. In contrast, inter-sectoral spillovers, as proxied by regional industrial diversity, reduce new firm formation across regions. The examination of other control variables suggests that GDP growth and small firms stimulate regional new firm formation, whereas sunk costs and unemployment have a discouraging effect.

Keywords

New firm formation knowledge spillovers regions manufacturing

Introduction

This research empirically examines the effect of knowledge spillovers on regional new firm formation for the Greek manufacturing sector over the period 2002–2010.

The theoretical framework that is particularly relevant to the present research is that of the knowledge spillover theory of entrepreneurship (KSTE) that was developed in a succession of research papers over the years (Acs et al., 2009, 2013; Audretsch and Lehmann, 2005). Within this framework, new knowledge is viewed as a source of entrepreneurial opportunities, and entrepreneurship itself works as a conduit that helps the commercialization and business application of the new knowledge. Knowledge externalities are crucial to the extent that the knowledge produced by existing firms and research organizations is not fully appropriated by them. Furthermore, entrepreneurship acts as a mechanism that further reinforces knowledge spillovers. These processes do not take place in a spatial vacuum. Spatial proximity to the knowledge source is of fundamental importance and knowledge spillovers are spatially bounded (Anselin et al., 1997; Bottazzi and Peri, 2003; Jaffe, 1989; Jaffe et al., 1993; Moreno et al., 2005; Sonn and Storper, 2008). According to Qian et al. (2013), the extent to which entrepreneurship facilitates knowledge spillovers depends on the absorptive capacity of regional entrepreneurship, and the quality of human capital is its key ingredient. On the other hand, Glaeser et al. (1992) identify two types of spatial industrial structure that are relevant to knowledge spillovers, one characterized by intra-sectoral spillovers due to spatial concentration of an industry (Marshall, 1920), and one characterized by inter-sectoral industry spillovers owing to the spatial proximity of diverse industries (Jacobs, 1969).

Audretsch and Lehmann (2005: p. 1193) vehemently argue that “while a large literature exists linking new-firm startup activity to region-specific characteristics and attributes…virtually none of these studies provided a theory linking knowledge spillovers to new-firm startup activity”. While things have changed since the emergence of KSTE and the studies have proliferated, the empirical evidence remains limited, especially in the case of less-advanced economies, and it is certainly missing for Greece.¹

In a Greek regional context, there have been empirical studies that have examined the effect of knowledge spillovers on the spatial agglomeration of manufacturing (Vogiatzoglou and Tsekeris, 2013) whereas Alexiadis and Tsagdis (2006) examined the factors that explain the location of R&D labor. Outside Greece, most studies have concentrated on the effects of knowledge spillovers on regional economic development and growth (Rodríguez-Pose and Crescenzi, 2008) and regional innovation (Anselin et al., 1997; Audretsch and Feldman, 1996). Although a number of authors have dealt with the effect of individual aspects of knowledge spillovers, based particularly on R&D and human capital, on new firm formation activity at a spatial level (Audretsch et al., 2010; Bania et al., 1993; Kirchhoff et al., 2007; Lasch et al., 2013; Woodward et al., 2006; Zucker et al., 1998), the literature that analyzes the relationship between knowledge spillovers and regional entrepreneurship is relatively unexplored (Alcacer and Chung, 2007; Audretsch et al., 2005; Audretsch and Keilbach, 2007; Audretsch and Lehmann, 2005).

The present paper differs from other empirical studies in that it uses a wider range of variables to proxy knowledge spillovers that are based on both innovation (e.g. R&D expenditures, R&D personnel, patent applications) and high-tech labor (e.g. scientists and engineers) in order to examine their influence on regional entrepreneurship. In addition, it considers both intra-sectoral and inter-sectoral types of knowledge spillovers that have been underlined in the literature (Doring and Schnellenbach, 2006; Van Stel and Nieuwenhuijsen, 2004). In this way, it attempts to shed more light on the relationship between knowledge spillovers and regional new firm formation rates.

The econometric analysis results obtained suggest that knowledge spillovers measured by indicators related to innovation and high-tech labor create conducive ground for the establishment of manufacturing new firms in Greek regions. Likewise, intra-sectoral spillovers are important drivers of regional entrepreneurial activity as geographic sectoral specialization and industrial intensity measures used reveal their substantial contribution to new firm formation levels. In contrast, inter-sectoral spillovers depicted by regional diversity lead to a reduction of manufacturing new firm formation across Greek regions.

The paper is organized as follows. The following section sets the theoretical background and motivation. The third section offers a description of the model variable construction and the sources of the data used. The fourth section presents the empirical framework along with the econometric methods employed, and the empirical results are described in the fifth section. The final section offers our concluding thoughts and policy implications.

Theoretical background

Knowledge spillovers

Knowledge spillovers are defined as “the external benefits from the creation of knowledge that accrue to parties other than the creator, occur at multiple levels of analysis, be it within or across organizations and networks” (Agarwal et al., 2010: p. 271). The absorption of knowledge from third-party firms takes place at no cost (Agarwal et al., 2010; Varga and Schalk, 2004). This means that the third-party firms do not (fully) compensate the firms that create knowledge for the benefits that they draw. In addition to the public character (Agarwal et al., 2010; Arrow, 1962) of knowledge spillovers, there exists a considerable amount of asymmetries, uncertainty and transactional costs (Agarwal et al., 2010; Arrow, 1962; Audretsch and Lehmann, 2005). Knowledge spillovers have been hard to measure and quantify (Krugman, 1991), and they have been associated with R&D spillovers (Jaffe, 1986) since R&D offers a means to measure knowledge inputs (Audretsch and Stephan, 1999). In turn, R&D spillovers may be either academic or industrial (Audretsch and Stephan, 1999; Carlsson et al., 2009). Apart, from R&D expenses which constitute an input in the innovation process, as an output measure, patent-related data have also been used (Bottazzi and Peri, 2003; Fritsch and Franke, 2004).

There has already been a growing literature on the positive impact of knowledge spillovers on regional economic growth (Rodríguez-Pose and Crescenzi, 2008; Varga and Schalk, 2004), on the reduction of regional productivity growth inequalities (Serrano and Cabrer, 2004), on regional human capital (Abel and Deitz, 2012) and on regional exports (Johansson and Karlsson, 2007).

A significant distinction between two types of knowledge spillovers, intra-sectoral and inter-sectoral spillovers, has been a dominant theme in the relative literature (Doring and Schnellenbach, 2006; Van Stel and Nieuwenhuijsen, 2004). The first relates to knowledge spillovers within the same industry, while the second pertains to knowledge spillovers across different industries. Intra-sectoral spillovers rely upon the MAR externalities theory (Arrow, 1962; Marshall, 1920; Romer, 1986, 1990). Marshall’s (1920) earlier consideration supports the view that the concentration of an industry in a particular location facilitates knowledge spillovers between firms and, consequently, the growth of that industry and that location. Arrow (1962) presents an early formalization of this, while Romer’s (1986, 1990) research is a more recent and influential statement which supports that this type of knowledge spillover is an important factor in the explanation of the differences in economic development across regions. In contrast, inter-sectoral spillovers correspond to the traditional Jacobs externalities notion (Jacobs, 1969). Jacobs (1969) notes that the most significant knowledge flows originate from outside the core industry where the firm operates. In this theory, cities constitute the spatial unit where the diversity of knowledge flows is maximized. This denotes that it is the geographical diversity of industries rather than geographical sectoral specialization that facilitates knowledge spillovers in addition to innovation and growth (Beaudry and Schiffauerova, 2009; Glaeser et al., 1992; Jacobs, 1969).

When it comes to the role of regional industrial diversity as a measure of inter-sectoral spillovers for regional economic growth and innovation, the evidence produced is not unequivocal as far as both positive (Feldman and Audretsch, 1999; Glaeser et al., 1992; Ouwersloot and Rietveld, 2000; Van Stel and Nieuwenhuijsen, 2004) and negative results have been produced (Henderson et al., 1995; Van der Panne, 2004).

The effect of knowledge spillovers on regional new firm formation

Knowledge spillovers create conducive ground to new entrepreneurship at a regional level. Reynolds et al. (1995: p. 391) hypothesize that “where information is readily available and innovation and creativity flourish, the formation rate of new firms is enhanced”. Bania et al. (1993) find that university research positively affects spatial new firm formation, especially in electrical and electronic equipment industries. In the same vein, Kirchhoff et al. (2007) show that university R&D expenditures act favorably towards regional new firm formation activity. Evidence pertaining to clusters of U.S. biotechnological firms suggests that these rely heavily on scientists with highly intellectual human capital (Zucker et al., 1998). In the case of Greece, Fotopoulos and Spence (1999), and Daskalopoulou and Liargovas (2010) find that human capital and skilled labor are critical determinants of regional new firm formation in manufacturing. In their study of German regions, Audretsch and Keilbach (2007) and Audretsch et al. (2010) examine the relationship between knowledge and new firm formation using the share of R&D workers, R&D intensity and the share of regional labor force accounted for by scientists and engineers as their preferred measures of knowledge. The evidence produced demonstrates that knowledge is indeed conducive to regional new firm formation. In Audretsch and Keilbach (2007) and Lasch et al. (2013), this is found to be especially so for knowledge-based industries, such as ICT and high-technology sectors.

Intra-sectoral spillovers appear to increase regional new firm formation rates in a number of studies. First, Lee et al. (2004) find that knowledge spillovers within the private sector encourage the regional new firm formation process. Acs and Armington (2004), and Acs et al. (2007) show that intra-sectoral spillovers are useful for the creation of new firms across regions in services, while Acs and Armington (2002), and Fotopoulos and Spence (1999) emphasize the value of intra-sectoral spillovers for regional new firm formation in manufacturing. Bosma et al. (2008) assess these results mentioning that the knowledge produced from firms of the same sector is the most conducive despite this knowledge being more limited as it refers to a specific sector (Bishop, 2012). Acs and Armington (2004), and Acs et al. (2007) claim that knowledge spillovers across different sectors have negative consequences on the regional formation of new firms in services. Specifically, Acs and Armington (2004) claim that spillovers present more benefits within industry sectors, but across sectors their role is rather negligible.

In this vein, many authors describe the positive role of geographic sectoral specialization in regional new firm formation and spin-off firms (Anyadike-Danes and Hart, 2006; Baltzopoulos et al., 2016; Coll-Martinez and Arauzo-Carod, 2017; Fotopoulos, 2014; Knoben et al., 2011). Anyadike-Danes and Hart (2006) support that business services specialization appears to have the largest contribution to new firm formation across regions of the United Kingdom, while Daskalopoulou and Liargovas (2010) additionally emphasize the benefits of specialization for Greek regional new firm formation in manufacturing.

It has been argued that “the higher the degree of diversification, the higher the variety of skills available locally. Skill and the diversity of working experiences can, in turn, give way to greater entrepreneurial choice and opportunity, especially when there is some degree of mobility of individuals between firms and industries” (Fotopoulos, 2014: p. 656). However, the results produced regarding the effect of regional industrial diversity on new firm formation have been rather contradictory. On the one hand there have been studies offering support for a positive effect of regional industrial diversity on regional entrepreneurship (Corradini and De Propris, 2015; Lasch et al., 2013; Rodríguez-Pose and Hardy, 2015), whereas others produce opposite results (Audretsch et al., 2010; Bishop, 2012; Fotopoulos, 2014). In their account of the results obtained in their study, Audretsch et al. (2010) explain that specialization externalities prevail over diversity. In turn, Bishop (2012) notes that the positive influences of diversity on new firm formation may be somewhat more sector-specific as they occur within the knowledge sector. In such a context, the concept of related variety² (Frenken et al., 2007) becomes most relevant. As exemplified by Boschma et al. (2013: p. 31), the concept of industry relatedness is positioned in the literature on spatial externalities and regional growth, and the related variety effect, “concerns externalities that come from a diversity of related industries in a region”. A positive effect of related variety on regional entrepreneurship is found in Bishop (2012), Baltzopoulos et al. (2016) and Bishop and Brand (2014). In the latter study, this positive effect applies to both manufacturing and services. On the other hand, unrelated variety does not appear to be a decisive factor for regional new firm formation in the Baltzopoulos et al. (2016) study and similarly Bishop and Brand (2014) specify that unrelated variety has an insignificant effect on new firm formation in manufacturing. Interestingly, however, a positive and significant impact for services was found in this study. Positive results of unrelated variety have also been produced by Bishop (2012).

Whatever the source of knowledge spillovers may be, it has been supported that these are geographically bounded (Anselin et al., 1997; Jaffe, 1989; Jaffe et al., 1993; Moreno et al., 2005) as the geographical proximity to the source of knowledge plays a key role. The localized nature of knowledge spillovers has also been very recently supported by Grillitsch and Nilsson (2017) and Roper et al. (2017). According to this view, geographical proximity to the source of knowledge plays a dominant role (Acs et al., 2013; Audretsch and Lehmann, 2005; Audretsch et al., 2005). Doring and Schnellenbach (2006) support that spatial proximity leads to lower costs of transmission and hence spillovers may be localized near to the source of knowledge. When the localized aspects of knowledge spillovers are considered, the distinction between codified and tacit knowledge becomes necessary (Acs and Varga, 2005; Audretsch and Lehmann, 2005). Tacit knowledge that pertains to partially codified and developed knowledge and is more sensitive to longer geographic distances and spatial proximity becomes more crucial (Audretsch and Feldman, 1996). Here, elements of social capital such as oral communication, personal contacts, trust and reciprocity facilitate the transfer of tacit knowledge (Acs and Varga, 2005; Audretsch and Stephan, 1996; Audretsch et al., 2005; Maskell and Malmberg, 1999). In contrast, codified knowledge is less sensitive to geographic distance. Indicatively, Varga and Schalk (2004) support that the transportation of codified knowledge over longer distances is materialized by patents and scientific articles. Thus, knowledge spillovers determine the location choice of firms (Alcacer and Chung, 2007; Audretsch and Lehmann, 2005; Audretsch et al., 2005).

The role of universities is crucial, as it has been argued that “universities in regions with a higher knowledge capacity and greater knowledge output also generate a higher number of technology startups” (Audretsch and Lehmann, 2005: p. 1201). Audretsch and Stephan (1996) argue that university-based scientists are responsible for the transfer of knowledge from scientific labs to firms as they can be involved in firms operating locally or take part in the establishment of new firms. Besides this, knowledge goes directly from university labs towards firms through: (a) personal contacts and face-to-face communication between firm employees and university-based scientists; and (b) firm employees who attend university seminars and workshops. This emphasizes the vital role of spatial proximity. Woodward et al. (2006) find that the positive effects of university R&D on regional high-tech new firm formation are limited to a geographic distance of 145 miles. In turn, Desrochers and Sautet (2008) support that firms of the same industry gain from spatial proximity as they exploit the common labor pool, better access to intermediate inputs, increased face-to-face communication and knowledge spillovers on related technologies. These authors, however, also emphasize the value of geographical proximity for firms of different sectors (diversified industries), underscoring that “face-to-face communication between individuals possessing different expertise is even more crucial than for experts in the same field (although the latter is typically considered very important) because they do not share a similar background and technical language. Face-to-face interaction was also deemed crucial to build trust relationships and to select suppliers” (Desrochers and Sautet, 2008: p. 824).

Based on the preceding discussion, the following key hypotheses are thus formed:

Hypothesis 1: Knowledge spillovers based on innovation and high-tech labor should positively affect regional new firm formation in the manufacturing sector.

Hypothesis 2: Intra-sectoral spillovers should have a positive impact on regional new firm formation rates in manufacturing.

Hypothesis 3: Inter-sectoral spillovers should have a significant effect on regional manufacturing new firm formation.

The direction of the latter effect cannot be a priori hypothesized as the results that have been previously produced in the relative literature have been mixed.

Data and variables

The data pertaining to the dependent and most of the independent variables, unless lagged, span the 2002 to 2010 period and pertain to 13 Greek Nomenclature of Territorial Units of Statistics (NUTS) 2 regions.³ In terms of the sector analyzed, this is the one-digit manufacturing sector aggregate.⁴ The regional distribution of manufacturing employment shares over the study period is presented in Table 1. The employment in manufacturing sector represents on average the 10% of the total regional employment for all the years of the model. The most industrialized regions are those of Kentriki Makedonia and Attiki, whereas the least industrialized one is that of the Ionia Nisia. By inspecting these employment shares across time, the existence of some temporal fluctuation also becomes evident.

Table 1.

Regional manufacturing employment shares.

	2002	2003	2004	2005	2006	2007	2008	2009	2010
Anatoliki Makedonia, Thraki	0.1254	0.1354	0.1322	0.1395	0.1238	0.1085	0.1110	0.0918	0.09
Kentriki Makedonia	0.1861	0.1874	0.1683	0.1684	0.1574	0.1611	0.1481	0.1321	0.13
Dytiki Makedonia	0.1160	0.0982	0.1227	0.1180	0.1361	0.1106	0.1102	0.1151	0.11
Ipeiros	0.0976	0.0988	0.101	0.0973	0.0890	0.0848	0.0896	0.0934	0.09
Thessalia	0.1215	0.1168	0.1184	0.1192	0.1293	0.1169	0.1222	0.1269	0.1
Ionia Nisia	0.0509	0.0379	0.03	0.0498	0.0492	0.0480	0.0443	0.0352	0.04
Dytiki Ellada	0.0936	0.0819	0.0853	0.0839	0.0856	0.0819	0.0798	0.0867	0.08
Sterea Ellada	0.1983	0.1707	0.1786	0.1635	0.1667	0.1610	0.1527	0.1660	0.18
Attiki	0.1572	0.1456	0.1456	0.1421	0.1397	0.1399	0.1316	0.1273	0.12
Peloponnisos	0.0791	0.0941	0.0897	0.0829	0.0867	0.0892	0.0703	0.0716	0.08
Voreio Aigaio	0.0758	0.0681	0.0732	0.0706	0.0795	0.0757	0.0719	0.0644	0.06
Notio Aigaio	0.0761	0.0885	0.0637	0.0648	0.0498	0.0532	0.0588	0.0605	0.06
Kriti	0.072	0.0787	0.0758	0.0732	0.0719	0.0772	0.0835	0.0792	0.07
Average	0.1115	0.1079	0.1065	0.1056	0.105	0.1006	0.098	0.0962	0.09

Average of regional share of manufacturing (2002–2010): 0.1025

Table 2.

Share of science and technology regional employment: 2002–2010.

	2002	2003	2004	2005	2006	2007	2008	2009	2010
Anatoliki Makedonia, Thraki	0.118	0.127	0.122	0.137	0.143	0.155	0.155	0.158	0.166
Kentriki Makedonia	0.181	0.18	0.196	0.195	0.201	0.208	0.214	0.213	0.211
Dytiki Makedonia	0.147	0.128	0.156	0.175	0.183	0.185	0.172	0.165	0.169
Ipeiros	0.166	0.167	0.169	0.183	0.185	0.183	0.183	0.17	0.164
Thessalia	0.149	0.165	0.15	0.161	0.197	0.187	0.205	0.216	0.180
Ionia Nisia	0.11	0.111	0.129	0.12	0.132	0.121	0.12	0.118	0.150
Dytiki Ellada	0.116	0.144	0.152	0.16	0.183	0.193	0.188	0.174	0.191
Sterea Ellada	0.085	0.105	0.134	0.142	0.143	0.128	0.13	0.14	0.138
Attiki	0.236	0.245	0.253	0.251	0.263	0.27	0.274	0.274	0.276
Peloponnisos	0.14	0.119	0.136	0.136	0.139	0.149	0.142	0.122	0.122
Voreio Aigaio	0.148	0.146	0.185	0.178	0.168	0.166	0.206	0.208	0.171
Notio Aigaio	0.102	0.115	0.108	0.14	0.15	0.142	0.139	0.109	0.106
Kriti	0.132	0.132	0.184	0.168	0.17	0.178	0.183	0.167	0.15
Average	0.141	0.145	0.16	0.165	0.174	0.174	0.178	0.172	0.169

Average of the ratio of employment in science and technology to workforce in Greek regions (2002–2010): 0.164

Table 3.

Patents to GDP ratio: Greek regions 2002–2010.

	2002	2003	2004	2005	2006	2007	2008	2009	2010
Anatoliki Makedonia, Thraki	0.0004	0.0004	0.00035	0.0325	0.0002	0.2844	0.1602	0.1079	0.0001
Kentriki Makedonia	0.5589	0.6433	0.3815	0.6794	0.7684	0.466	0.3602	0.5035	0.5125
Dytiki Makedonia	0.0003	0.0003	0.0003	0.4404	0.0003	0.0001	0.0002	0.2087	0.4227
Ipeiros	0.0886	0.5599	0.00035	0.1602	0.3017	0.0003	0.4382	0.204	0.0003
Thessalia	0.4661	0.0256	0.3906	0.4839	0.1994	0.5929	0.1247	0.0856	0.1117
Ionia Nisia	0.0003	0.0002	0.2852	0.0002	0.0001	0.2399	0.0002	0.0001	0.0002
Dytiki Ellada	0.4331	0.5823	0.1674	0.4364	0.5911	0.9326	0.1099	0.3631	0.2606
Sterea Ellada	0.3156	0.0059	0.1738	0.2429	0.0002	0.1032	0.0581	0.3831	0.2403
Attiki	0.5882	0.6083	0.4538	0.6762	0.5386	0.5071	0.442	0.4194	0.321
Peloponnisos	0.0367	0.4106	0.00034	0.2607	0.2693	0.2831	0.4674	0.000279	0.1151
Voreio Aigaio	0.4885	0.0004	0.0003	0.0003	0.0003	0.0003	0.0003	0.0003	0.0003
Notio Aigaio	0.0003	0.0002	0.0754	0.1412	0.2597	0.0096	0.5612	0.2005	0.0002
Kriti	0.7354	0.6138	0.4166	0.7031	0.7763	0.2988	0.6062	0.6479	0.1626
Average	0.2856	0.2655	0.1805	0.3275	0.2851	0.2861	0.2561	0.2403	0.1652

Average of the ratio of patents to GDP in Greek regions (2002–2010): 0.2546; GDP in million euro.

This helps to give some background information in relation to the varying degrees of industrialization across Greek regions.⁵

Dependent variable

The variable of interest and dependent variable of this study is regional manufacturing new firm formation rates (NFFR). There have been two approaches in calculating regional new firm formation rates (Audretsch and Fritsch, 1994). The first, termed as the “ecological” approach, standardizes the number of new firms in a region by dividing them by the number of firms in existence at the beginning of the period considered (in this case, the corresponding year each time), while the second standardizes the number of new firms with respect to the size of regional work force. Garofoli (1994) criticizes the “ecological” approach on the grounds that this essentially assumes that new firms emanate from the stock of existing firms. It is quite often the case that a very high correlation exists between the numerator (new firms) and the denominator (number of existing firms). Moreover, Garofoli (1994) stresses that the “ecological” approach does not take into account the difference in firm sizes between the numerator and the denominator. The resulting new firm formation rate could be biased upward in regions dominated by fewer larger firms.

On the other hand, the labor force appears to accord with the fact that it is ultimately real people who establish firms. Fotopoulos and Spence (1999: p. 221) further argue that “the main attraction is that a potential entrepreneur aiming to open a business often draws on previously gained experience as an employee in the same labor market area within which the new establishment is due to become operational”.

The present analysis thus follows the labor force approach⁶ in order to express new firm formation rates. These rates refer to the ratio of the number of new firms in each region for manufacturing over the regional employment in manufacturing sector. The data for the number of new firms and employment for the construction of new firm formation rates come from the Firm’s Registry and Labor Force Survey (LFS) of the Hellenic Statistical Authority (EL.STAT.)

Figure 1 shows the Greek regional distribution of new firm formation rates. There, the highest new firm formation rates are observed in the most rural and unindustrialized areas (e.g. Dytiki Makedonia, Ipeiros, Ionia Nisia, Voreio Aigaio, and Notio Aigaio). Similar patterns have also been found in a Greek regional (Fotopoulos and Spence, 1999) and international context (Acs and Armington, 2002; Gould and Keeble, 1984; Keeble and Walker, 1994; Lee et al., 2004).

Figure 1.

Regional distribution of new firm formation rates: 2002–2010 period averages.

Independent variables

Knowledge spillovers are measured through the use of a number of proxies. The first proxy of knowledge spillovers is manufacturing R&D expenditure intensity ( RDEXP ). Following Bottazzi and Peri (2003) this is defined as R&D expenditures per square kilometer. As the knowledge spillover effect is very localized, this essentially R&D density measure is thought to reflect this better. The data for the manufacturing R&D expenditure is drawn from the Annual Industrial Survey of the EL.STAT. Along with manufacturing R&D, the present analysis also uses public (government and higher education) sector R&D expenditures per square kilometer ( RDPUB ).⁷ The data required for the construction of the public sector R&D expenditure were derived from Eurostat’s regional database.

Following Audretsch et al. (2010), data on R&D personnel have also been used. Here again the R&D personnel number has been expressed in per square kilometer terms to result in an R&D personnel density variable ( RDPERS ). The rationale is that higher R&D personnel densities facilitate sharing of information and knowledge through personal interaction and, through this, positively affect regional new firm formation rates. Manufacturing R&D personnel intended for the construction of R&D personnel intensity originates from the Annual Industrial Survey of the EL.STAT.

Moving from the inputs to the innovation and knowledge creation process to its output metrics, patent intensity ( PATINT ) is employed as another measure of knowledge spillovers. Patent intensity is defined as the number of patents in each region over the number of regional inhabitants (Carlino et al., 2007; Goncalves and Almeida, 2009). The data used for the construction of this variable are taken from Eurostat.

Two additional variables accounting for the density of scientists and engineers per square kilometer⁸ ( SCIENGIN ) and the density of employment in science and technology sectors per square kilometer ( EMPSCITECH ), were also used. In a similar vein, Bishop (2012) chooses employment in two-digit knowledge intensive sectors and high-tech manufacturing as a proxy of regional knowledge stock. In both cases, the employment numbers come from Eurostat.

Figures 2, 3, 4, and 5 detail the regional distribution of patent intensity, manufacturing R&D expenditures intensity, R&D personnel, and scientist-and-engineer employment intensities. The geographical distribution of all measures exhibits a strong concentration in the Attiki and Kentriki Makedonia regions and is quite revealing about the extent of spatial heterogeneity that these measures exhibit.

Figure 2.

Regional distribution of patent intensity: 2002–2010 period averages.

Figure 3.

Regional distribution of manufacturing R&D expenditures intensity: 2002–2010 period averages.

Figure 4.

Regional distribution of R&D personnel intensity: 2002–2010 period averages.

Figure 5.

Regional distribution of scientists and engineers intensity: 2002–2010 period averages.

Intra-sectoral spillovers are measured by: (a) industrial intensity; and (b) the Hoover specialization index. The industrial intensity variable ( INDINTENS1 ) is defined as the number of existing manufacturing establishments per capita (Acs et al., 2007; Acs and Armington, 2002, 2004). An additional measure of industrial intensity ( INDINTENS2 ), essentially an industrial density measure, was also used, defined as the number of manufacturing establishments per square kilometer (Cainelli et al., 2016; Ganau, 2016). The data required for the construction of these variables come from the Firm’s Registry and the Population Census of the EL.STAT.

As discussed earlier, regional sectoral-specialization has been often utilized as an additional proxy for intra-sectoral spillovers (Van Stel and Nieuwenhuijsen, 2004). In view of this, the present research utilizes the Hoover specialization ( HOOVIND ) index ( Hoover, 1941; see also Fotopoulos and Spence, 1999) defined as

H O O V I N D = \frac{1}{2} * \sum_{i} | \frac{E_{i r}}{E_{r}} - \frac{E_{i n}}{E_{n}} |

(1)

where E_ir is the employment in industry

i = 1, \dots, J

and

r = 1, \dots, R

regions at some point in time, E_in is the employment in industry

i

nationally, while E_r and E_n pertain to total manufacturing employment at the regional and national level, respectively.

The data for the construction of the Hoover index come from the LFS. Following Van Stel and Nieuwenhuijsen (2004) who consider the overall regional industrial diversity as a proxy of inter-sectoral knowledge spillovers, a regional diversity index was used by the present research to capture the effect of such spillovers. This is the entropy-based Theil ( THEIL ) index (Theil, 1972: p. 26; see also Fotopoulos, 2014; Rodríguez-Pose and Hardy, 2015) defined as

T_{r} = \sum_{i} \frac{p_{r i}}{p_{r}} log \frac{p_{r}}{p_{r i}}

(2)

where E_ri is the employment in region

r = 1, \dots., R

and industry

i = 1, \dots, J

p_{r i} = \frac{E_{r i}}{\sum_{r} \sum_{i} E_{r i}}

, and

p_{r} = \sum_{i} p_{r i}

This measure takes the value of 0 when only one sector is present in region r and the value $\ln (n)$ where all $n$ two-digit industrial sectors use the same number of persons in the region in question. The source of data for the construction of the Theil index is LFS.

The definition of unrelated ( UNRELATED ) and related variety ( RELATED ) is based on Frenken et al. (2007).⁹ They define an entropy-based measure of unrelated variety as

U V = \sum_{g = 1}^{G} P_{g} {log}_{2} (\frac{1}{P_{g}})

(3)

In our analysis, all two-digit industries $i$ are part of a one-digit sector $S_{g}$ , for $g = 1, \dots, G$ . Then, one-digit shares, $P_{g}$ , arise from summing up the two-digit employment shares $p_{i}$ such that

P_{g} = \sum_{i \in S_{g}} p_{i}

(4)

In turn, related variety is calculated as the weighted sum of entropy within each one-digit sector as

R V = \sum_{g = 1}^{G} P_{g} H_{g}

(5)

where

H_{g} = \sum_{i \in S_{g}} \frac{P_{i}}{P_{g}} log (\frac{1}{\frac{P_{i}}{P_{g}}})

(6)

The sum of related and unrelated variety gives the total entropy index (Theil, 1972: p. 20–22; see also Frenken et al., 2007) calculated in (2) such that

T E = U V + R V

(7)

Control variables

A sunk costs ( SUNK ) variable at the regional level, defined as 1 minus the ratio of second-hand bought machinery and equipment over the total investments in machinery and equipment (see Fotopoulos and Louri, 2000), was used by utilizing data from the Annual Industrial Survey. Sunk costs are past costs that have already been incurred and cannot be recovered. In a spatial context, Stam (2007) notes that there may be physical investments that are tied to the current location or, at least, to the current region. Both theorists in mainstream economics (Caves and Porter, 1977; Eaton and Lipsey, 1980, 1981) and in economic geography (Clark and Wrigley, 1997a, 1997b) have viewed sunk costs as significant barriers to entry. As a result, our analysis expects a negative impact of sunk costs on regional manufacturing new firm formation.

To account for the effect of wider economic conditions, the analysis first considers the effect of the regional unemployment rate (UNEMP) averaged over the previous three years. This is a variable that has produced both positive (Audretsch and Fritsch, 1994) and negative results (Acs and Armington, 2002; Lee et al., 2004) when the analysis of regional new firm formation rates concentrates on manufacturing. Next, the effect of local demand growth as proxied by GDP growth is examined. GDP growth ( GDPGR ) refers to the growth of GDP over the three preceding years, each time. A number of studies find a positive influence of GDP growth on regional new firm formation in manufacturing (Fotopoulos and Spence, 1999). Others have produced an insignificant effect (Acs and Armington, 2002; Reynolds et al., 1994), while in Lee et al. (2004) their income growth variable has been found positive and significant, save for the case of manufacturing industries.

The last control variable considered is one with a long history in regional new firm formation studies, (for a recent discussion see Fotopoulos, 2014), that which accounts for the extent of small firm presence regionally. The corresponding variable ( SMFP ) is defined as the ration of the number of establishments with less than ten employees divided by the total number of existing establishments. The number of small firms locally can reflect a larger local supply of role models for aspiring entrepreneurs, a larger local supply of employees who, by working for small firms, have acquainted themselves with a wider range of everyday business activities and are thus more prepared to undertake their entrepreneurial initiative, and possibly also a lower regional level of entry barriers. For these reasons, SMFP is expected to have a positive effect on regional new-manufacturing-firm formation rates.

Empirical framework

As the data in hand have both a cross sectional (regions) and a time (years) dimension, the variables employed in the analysis can have two systematic components of variation: one that pertains to differences between regions and one that pertains to differences between years. As has been suggested by Cameron and Trivedi (2005: p. 709), “with panel data it is useful to know whether variability is across individuals [regions in our case] or across time”. This is important to the extent that the coefficients obtained from fixed-effects estimation of time-variant variables “can be very imprecise if most of the variation in a regression is cross sectional rather than over time” (Cameron and Trivedi, 2005: p. 715).

Table 4 presents the results of a variance decomposition exercise for both the dependent variable and some key independent variables, (for technical details see Körösi et al., 1992: p. 194-197; Fotopoulos and Spence, 1999).

Table 4.

Variance decomposition.

Variable	Between-regions/ $σ^{2}$	Between-time/ $σ^{2}$	Residual/ $σ^{2}$	Systematic(model)/ $σ^{2}$	F-region	F-time
NFFR_labour market approach	0.644	0.2048	0.1517	0.8484	33.95 (0.0)	16.2 (0.0)
NFFR “ecological” approach	0.137	0.6879	0.1754	0.8246	6.24 (0.0)	47.07 (0.0)
INDINTENS1	0.889	0.047	0.0636	0.9364	111.93 (0.0)	8.864 (0.0)
HOOVIND	0.769	0.0441	0.1872	0.8128	32.86 (0.0)	2.83 (0.0)
PATINT	0.643	0.0402	0.3163	0.6837	16.27 (0.0)	1.524 (0.2)
THEIL	0.595	0.3537	0.0512	0.9488	93.1(0.0)	82.97 (0.0)
RDPUB	0.952	0.0045	0.0431	0.96	177.01 (0.0)	1.253 (0.3)
INDINTENS2	0.975	0.0027	0.0223	0.9777	349.5 (0.0)	1.467 (0.2)
RDEXP	0.959	0.0029	0.0381	0.9619	201.3 (0.0)	0.927 (0.5)
SCIENGIN	0.987	0.0018	0.0109	0.9891	723.63 (0.0)	1.975 (0.06)
RDPERS	0.851	0.0119	0.1367	0.8633	49.84 (0.0)	1.048 (0.4)

F-test for regional and time effects (probability of F in parentheses).

Taking the new firm formation rate (NFFR) variables, the first striking observation is that in the one based on the labor market approach, the between-regions variation accounts for 64% of the overall variation in this variable, whereas the corresponding value for the between variation in the case of the “ecological” version of the NFFR is just about 14%. In contrast, it appears that 68% of the overall variation in the “ecological” NFFR is accounted by between-year variation.

When it comes to independent variables relating one way or another to knowledge spillovers, in all cases the between-region variation represents the largest, by far, of the systematic components. The only variable for which the between-time variation represents a sizeable proportion (35%) of the overall variation is that of the THEIL diversity index.

As we are interested in accounting for the effect of knowledge spillovers and other regional control variables on new firm formation, these decomposition results leave us with little choice and essentially dictate the adoption of the labor-market-approach-based NFFR.

With the between-time variation being the dominant systematic source of variation in both the dependent and independent variables the regional fixed-effects estimator that would account for unobserved regional heterogeneity will be highly problematic. As Beck (2001: p. 285) very aptly describes, “with slowly changing independent variables, the fixed effects will soak up most of the explanatory power of those slowly changing variables… [and] will make it hard for such variables to appear either substantially or statistically significant”.

Given the results of the variance decomposition exercises and the concerns that have been expressed in the literature regarding the appropriateness of cross-sectional fixed effects estimators, we resort to pooled estimators with heteroscedasticity robust standard errors. The pooled estimator has often been the choice in cases of time-invariant and relatively time-invariant independent variables ( Oaxaca and Geisler, 2003; see also Knack, 1993). The basic pooled ordinary least squares (OLS) model is expressed as

\begin{array}{l} N F F R_{r t} = a + β_{1} P A T I N T_{r t - 1} + β_{2} R D E X P_{r t - 1} + β_{3} R D P U B_{r t - 1} + β_{4} R D P E R S_{r t - 1} \\ + β_{5} S C I E N G I N_{r t - 1} + β_{6} E M P S C I T E C H_{r t - 1} + β_{7} I N D I N T E N S 1_{r t - 1} \\ + β_{8} I N D I N T E N S 2_{r t - 1} + β_{9} H O O V I N D_{r t - 1} + β_{10} T H E I L_{r t - 1} \\ + β_{11} R E L A T E D_{r t - 1} + β_{12} U N R E L A T E D_{r t - 1} + γ_{1} S U N K_{r t - 1} \\ + γ_{2} G D P G R_{r t - 1} + γ_{3} U N E M P_{r t - 1} + γ_{4} S M F P_{r t - 1} + e_{r t} \end{array}

(8)

where

N F F R

is the dependent variable, and the variables on right-hand side are the independent and control variables as described previously. In turn,

a

is the constant term and

β = (β_{1}, \dots, β_{12})

and

γ = (γ_{1, \dots,} γ_{4})

are the vectors of coefficients of independent and control variables, respectively. On the other hand,

r

and

t

are the indices for regions (

r = 1, 2, \dots, 13)

and time (

t = 1, 2, \dots, 9)

, while

e

is the error term. The model introduces independent and control variables with one lag, an approach that has often been used in order to avoid possible reverse causality and endogeneity problems (see for example Clemens et al., 2012; Green et al., 2005).

As 20% of the overall variation in the dependent variable is accounted for by between-year variation, a time fixed-effects estimation was also performed.

The time-effects model is described by

\begin{array}{l} N F F R_{r t} = a + β_{1} P A T I N T_{r t - 1} + β_{2} R D E X P_{r t - 1} + β_{3} R D P U B_{r t - 1} + β_{4} R D P E R S_{r t - 1} \\ + β_{5} S C I E N G I N_{r t - 1} + β_{6} E M P S C I T E C H_{r t - 1} + β_{7} I N D I N T E N S 1_{r t - 1} \\ + β_{8} I N D I N T E N S 2_{r t - 1} + β_{9} H O O V I N D_{r t - 1} + β_{10} T H E I L_{r t - 1} \\ + β_{11} R E L A T E D_{r t - 1} + β_{12} U N R E L A T E D_{r t - 1} + γ_{1} S U N K_{r t - 1} \\ + γ_{2} G D P G R_{r t - 1} + γ_{3} U N E M P_{r t - 1} + γ_{4} S M F P_{r t - 1} + δ_{t} + e_{r t} \end{array}

(9)

where

δ_{t}

is the time effect. The latter term distinguishes equation (9) from equation (8).

Empirical results

Before getting to the actual estimation, Table 5 presents some basic descriptive statistics of all variables in the model while Table 6 gives the correlation coefficient matrix for the explanatory variables involved in the econometric analyses to follow.

Table 5.

Descriptive statistics (N = 117).

Variables	Mean	Std. deviation	Min	Max
1. NFFR	0.0105	0.0053	0.002	0.029
2. UNEMP	0.1039	0.0203	0.065	0.1656
3. GDPGR	0.0637	0.0168	0.003	0.1019
4. HOOVIND	0.0339	0.0106	0.018	0.0707
5. INDINTENS1	0.008	0.0022	0.0051	0.0169
6. SCIENGIN	0.0026	0.0068	0.0001	0.0289
7. PATINT	4.36	4.55	0.0043	16.91
8. SUNK	0.7593	0.3506	0.0029	0.8904
9.EMPSCITECH	0.0134	0.0333	0.0013	0.1385
10.THEIL	4.2717	0.3178	3.7165	5.2268
11. RELATED	0.839	0.2395	0.48	1.4919
12.UNRELATED	3.4326	0.1203	3.159	3.7489
13. INDINTENS2	1.0547	2.1536	0.2137	9.7537
14. SMFP	0.949	0.0259	0.88	0.985
15. RDEXP	1.0838	3.5947	0.0001	16.47
16. RDPUB	1.1529	3.2644	0.001	16.63
17. RDPERS	0.0183	0.0615	0.0015	0.3105

Table 6.

Correlation matrix with correlation coefficients of variables (N = 117).

Variables	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.	13.	14.	15.	16.	17.
1. NFFR	1.00
2. UNEMP	0.14	1.00
3. GDPGR	0.26	−0.24	1.00
4. HOOVIND	0.43	0.47	−0.11	1.00
5. INDINTENS1	0.28	0.54	−0.02	0.38	1.00
6. SCIENGIN	−0.3	−0.2	0.27	−0.4	0.009	1.00
7. PATINT	−0.3	−0.3	0.14	−0.4	0.06	0.58	1.00
8. SUNK	0.29	0.21	0.16	0.09	0.17	−0.0	0.059	1.00
9. EMPSCITECH	−0.3	−0.2	0.28	−0.4	0.01	0.99	0.58	−0.0	1.00
10. THEIL	−0.6	−0.06	−0.02	−0.2	0.07	0.56	0.44	−0.5	0.56	1.00
11. RELATED	−0.7	−0.14	−0.08	−0.3	−0.1	0.57	0.46	−0.4	0.57	0.95	1.00
12. UNRELATED	−0.2	0.11	0.097	0.07	0.29	0.35	0.24	−0.5	0.34	0.76	0.51	1.00
13. INDINTENS2	−0.3	−0.15	0.28	−0.4	0.08	0.97	0.59	0.05	0.97	0.51	0.54	0.28	1.00
14. SMFP	0.74	0.199	0.03	0.5	0.22	−0.6	−0.43	0.22	−0.6	−0.7	−0.8	−0.3	−0.5	1.00
15. RDEXP	−0.3	−0.1	0.25	−0.4	0.01	0.97	0.55	−0.0	0.97	0.54	0.55	0.32	0.96	−0.5	1.00
16. RDPUB	−0.3	−0.2	0.27	−0.4	0.003	0.99	0.61	−0.0	0.99	0.57	0.57	0.36	0.94	−0.6	0.93	1.00
17. RDPERS	−0.3	−0.1	0.24	−0.4	0.06	0.88	0.52	0.07	0.89	0.46	0.5	0.23	0.95	−0.5	0.92	0.83	1.00

Some pairs of explanatory variables present a high degree of correlation. The following correlation coefficients stand out: (a) patent intensity and R&D expenditures intensity (0.55); (b) patent intensity and R&D expenditures intensity in public sector (0.61); (c) R&D expenditures intensity and R&D expenditures intensity in public sector (0.93); (d) industrial intensity defined as the number of manufacturing establishments per square kilometer and R&D expenditures intensity (0.96); (e) Theil index (overall variety) and related variety (0.95); (f) Theil index and unrelated variety (0.76); and (g) related and unrelated variety (0.51). As a result, and to avoid multicollinearity, variables that proxy similar economics concepts were kept separate in the different econometric estimation permutations. As an additional measure, the variance inflation factor (VIF) was also used to detect any possible remaining multicollinearity problems. The VIF values of 10 but also of 5 have been suggested in the literature as critical cutoff points above which multicollinearity emerges as a problem (see Kutner et al., 2004; Rogerson, 2001). The average values of VIF for the variables of the model are given for econometric-estimation permutation columns in the results that follow.

The econometric analysis starts off with a pooled-OLS estimation as the regional fixed-effects approach is ruled out for the reasons discussed earlier. The results of the pooled-OLS estimation are presented in Table 7, whereas the results of the time fixed-effects estimation are presented in Table 8. Both tables display the estimated coefficients along with robust standard errors in parentheses.

Table 7.

Pooled-OLS estimations (N = 117) dependent variable: new firm formation rate.

Variables	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.
RDPUB	−	−	−	0.0003***	−	−	−	−	−	−
				(0.0000)
PATINT	−	−	0.0002**	−	−	−	−	−	−
			(0.0000)
RDEXP	0.0002**	0.0002***	−	−	−	−	−	−	−	−
	(0.0000)	(0.0000)
RDPERS	−	−	−	−	0.0068*	0.0093**	−	−	−
					(0.0042)	(0.0042)
SCIENGIN	−	−	−	−	−	−	−	0.0843*	0.0531	−
								(0.0504)	(0.052)
EMPSCITECH	−	−	−	−	−	−	−	−	−	0.0178*
										(0.0104)
HOOVIND	0.0698**	0.0653**	0.1618***	0.0618**	0.0668*	0.0651**	−	−	0.0395	−
	(0.0362)	(0.0303)	(0.037)	(0.0294)	(0.0366)	(0.0311)			(0.0398)
SUNK	−0.0001	−0.0002	−0.0023	−6.02e-07	−3.57e-06	−0.0002	−0.0003	−0.00005	0.0016	−0.00007
	(0.0013)	(0.0012)	(0.0014)	(0.0011)	(0.0014)	(0.0012)	(0.0012)	(0.0013)	(0.0014)	(0.0013)
THEIL	−0.005***	−	−0.01***	−	−0.004***	−	−	−0.004***	−	−0.004***
	(0.0018)		(0.0014)		(0.0018)			(0.0017)		(0.0017)
SMFP	0.1073***	0.0902***	−	0.094***	0.1066***	0.0881***	0.1089***	0.1252***	0.1483***	0.1252***
	(0.0193)	(0.022)	(0.022)	(0.0193)	(0.0222)	(0.0194)	(0.0195)	(0.0177)	(0.0195)
UNEMP	0.0133	0.0111	0.0317*	0.0139	0.0127	0.0109	0.0247	0.0115	0.0057	0.0117
	(0.0153)	(0.015)	(0.0191)	(0.0152)	(0.0151)	(0.0149)	(0.0153)	(0.0172)	(0.0155)	(0.0172)
GDPGR	0.0783***	0.071***	0.0999***	0.0705***	0.0809***	0.0737***	0.0705***	0.0758***	0.0741***	0.0752***
	(0.0144)	(0.0147)	(0.0165)	(0.0145)	(0.0144)	(0.0145)	(0.0148)	(0.0152)	(0.015)	(0.0152)
INDINTENS1	−	−	−	−	−	−	−	0.2632*	−	0.2618*
								(0.1622)		(0.1621)
INDINTENS2	−	−	−	−	−	−	0.0004***	−	−	−
							(0.0002)
RELATED	−	−0.008***	−	−0.008***	−	−0.008***	−0.008***	−	−	−
		(0.0025)		(0.0025)		(0.0025)	(0.0023)
UNRELATED	−	−	−	−	−	−	−	−	−0.0006	−
									(0.0037)
CONS	−0.081***	−0.077***	0.0434***	−0.08***	−0.08***	−0.075***	−0.094***	−0.099***	−0.14***	−0.099***
	(0.0233)	(0.0218)	(0.0059)	(0.0218)	(0.0231)	(0.022)	(0.0191)	(0.0223)	(0.0216)	(0.0223)
R ²	0.6432	0.6565	0.5746	0.6575	0.6397	0.6534	0.6510	0.6440	0.6267	0.6445
F	30.91	32.92	27.72	33.50	30.51	31.39	36.53	28.57	30.56	28.7
Avg VIF	2.46	2.48	1.76	2.29	2.3	2.34	2.22	2.11	1.67	2.11

Significant at 10%; ** significant at 5%; ***significant at 1%; VIF: variance inflation factor

Table 8.

Time effects estimations (N = 117) dependent variable new firm formation rate.

Variables	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.
RDPUB	−	−	−	0.0003***	−	−	−	−	−	−
				(0.0000)
PATINT	−		0.0001*	−	−	−	−	−	−	−
			(0.0000)
RDEXP	0.0002**	0.0002***	−	−	−	−	−	−	−	−
	(0.0001)	(0.0000)
RDPERS	−	−	−	−	0.0081*	0.0089**	−	−	−	−
					(0.0049)	(0.0045)
SCIENGIN	−	−	−	−	−	−	−	0.0965*	0.0857	−
								(0.0526)	(0.0589)
EMPSCITECH	−	−	−	−	−	−	−	−	−	0.021**
										(0.0111)
HOOVIND	0.0701*	0.0693**	0.1584***	0.0661**	0.0639	0.0668*	−	−	0.0393	−
	(0.0404)	(0.0353)	(0.0391)	(0.0343) (0.0409)	(0.0359)	(0.0441)
SUNK	0.0016	0.0006	−0.0014	0.0004	0.0013	0.0003	−0.0000	0.0004	0.0018	0.0004
	(0.0025)	(0.0026)	(0.0026)	(0.0026)	(0.0025)	(0.0025)	(0.0025)	(0.0025)	(0.0025)	(0.0025)
THEIL	−0.0049 ***	−	−0.011***	−	−0.004**	−	−	−0.004**	−	−0.0041**
	(0.0021)		(0.0016)		(0.002)			(0.0018)		(0.0018)
SMFP	0.1169***	0.0863***	−	0.0895***	0.115***	0.0835***	0.1104***	0.1323***	0.1579***	0.1328***
	(0.0202)	(0.0275)		(0.0276)	(0.02)	(0.0278)	(0.0233)	(0.0218)	(0.0186)	(0.0218)
UNEMP	0.0148	0.0083	0.0326*	0.0109	0.0146	0.0091	0.0237	0.009	0.0089	0.0097
	(0.0143)	(0.0137)	(0.0174)	(0.0136)	(0.0141)	(0.0136)	(0.0128)	(0.0168)	(0.0166)	(0.0169)
GDPGR	0.0725***	0.0641***	0.0998***	0.0622***	0.0768***	0.069***	0.0642***	0.0681***	0.0657***	0.0674***
	(0.0152)	(0.0146)	(0.0164)	(0.0147)	(0.0146)	(0.0137)	(0.0148)	(0.0168)	(0.0181)	(0.0169)
INDINTENS1	−	−	−	−	−	−	−	0.2815*	−	0.2762*
								(0.1675)		(0.1679)
INDINTENS2	−	−	−	−	−	−	0.0004***(0.0002)	−	−	−
RELATED	−	−0.0089***	−	−0.0088***	−	−0.008***	−0.008***	−	−	−
		(0.0032)		(0.0031)		(0.0032)	(0.0028)
UNRELATED	−	−	−	−	−	−	−	−	−0.0017	−
								(0.0046)
CONS	−0.091***	−0.0725***	0.0421***	−0.0754***	−0.092***	−0.07***	−0.095***	−0.11***	−0.14***	−0.1069***
	(0.0243)	(0.0279)	(0.0058)	(0.028)	(0.0242)	(0.0283)	(0.024)	(0.0251)	(0.0215)	(0.0251)
R ²	0.6905	0.7026	0.6195	0.7055	0.6851	0.698	0.6972	0.6927	0.6767	0.6935
F	23.87	25.40	18.29	27.19	23.46	24.46	25.72	22.25	23.57	22.16
F-time effects	F₁(8,101) = 1.9	F₂(8,101) = 1.96	F₃(8,102) = 1.5	F₄(8,101) = 2.1	F₅(8,101) = 1.8	F₆(8,101) = 1.9	F₇(8,102) = 1.9	F₈(8,101) = 2	F₉(8,101) = 1.96	F₁₀(8,101) = 2

Significant at 10%; ** significant at 5%; ***significant at 1%

The results that cover the period from 2002 to 2010¹⁰ suggest that knowledge spillovers as proxied in the present research form a conducive ground for the establishment of manufacturing new firms across Greek regions. Both innovation-related (e.g. patent intensity, R&D expenditures intensity, R&D expenditures intensity in public sector, R&D personnel intensity) and science and technology employment intensity measures (e.g. scientist-and-engineer intensity, employment in science and technology intensity) are also found to be conducive to regional entrepreneurship. These findings accord with the results of previous studies offering evidence for the beneficial effect of knowledge spillovers on regional new firm formation (Audretsch et al., 2010; Audretsch and Keilbach, 2007; Audretsch and Lehmann, 2005; Kirchhoff et al., 2007).

When it comes to the distinction between intra-sectoral and inter-sectoral spillovers, the former appears to have a positive and significant impact on regional new firm formation rates. This holds for all the intra-sectoral spillovers measures employed in the present analysis. This finding essentially suggests that the benefits of geographic sectoral specialization encourage new firm formation. This result was obtained in both the OLS and time effects models and is in line with those authors who make references to similar findings (Acs and Armington, 2004; Baltzopoulos et al., 2016; Knoben et al., 2011; Lee et al., 2004).

In contrast, inter-sectoral spillovers, as measured by the overall regional industrial diversity (Theil index), were found to have a negative effect on regional entrepreneurial activity. Related and unrelated variety have been used as additional measures of inter-sectoral knowledge spillovers. The related variety measure was found to have a negative effect on regional new firm formation whereas the effect of unrelated variety was statistically insignificant. These results concur with prior research that supports that knowledge spillovers across sectors tend to decrease regional entrepreneurship rates (Acs et al., 2007; Acs and Armington, 2004; Fotopoulos, 2014) whereas they clearly contradict some other findings (Corradini and De Propris, 2015; Rodríguez-Pose and Hardy, 2015). As opposed to regional sectoral specialization, regional industrial diversity has a negative effect on regional new firm formation in manufacturing. This result can be reconciled with Fotopoulos and Spence’s (1999) earlier findings for Greece on the positive effect of production specialization on regional new firm formation.

As far as the effect of other control variables is concerned, GDP growth is found to operate as a significant driver of regional new firm formation employment rates. This result points to the importance of demand-side factors when it comes to creating positive business prospects and through that encouraging new firm formation. On the other hand, the effect of regional unemployment was found to be statistically insignificant in both pooled-OLS and time-fixed effects estimation results.

The extent of small firm presence regionally has been found to have a positive and significant effect on new firm formation, a result that concurs with earlier findings in both Greek (Fotopoulos and Spence, 1999) and other country contexts (see Reynolds et al., 1994 for some early cross-national evidence). Finally, the results on the regional sunk cost proxy was not found to have an effect that is statistically different from zero.

Conclusions

The present paper examines the effect of knowledge spillovers on new firm formation across Greek regions in the manufacturing sector. It employs different measures of knowledge spillovers based on innovation and “high-tech” labor as well as differentiating between intra-sectoral and inter-sectoral spillovers. The distinguishing feature of the paper is that it offers one of the most detailed accounts of knowledge-spillover related variables in manufacturing new firm formation. It also adds to the small but growing literature by offering evidence from a less advanced and recently troubled economy.

The results obtained suggest that overall knowledge spillovers seem to have a positive effect on regional new firm formation rates in Greece. This has been corroborated by the results obtained from different variables used to proxy knowledge spillovers such as patents intensity, R&D expenditures intensity, R&D personnel intensity and high-tech labor measures, all found to be positive and significant in different model estimation permutations. Our findings, however, also suggest that the relevant knowledge spillover concept when it comes to entrepreneurship rates in Greek regions is that of intra-sectoral rather than inter-sectoral spillover. Regional sectoral specialization is more conducive to new firm formation rather than sectoral diversity and variety measures. The significant effect of intra-sectoral spillovers on regional new firm formation has been found to apply not only in manufacturing (Acs and Armington, 2002; Fotopoulos and Spence, 1999) but also in the services sector (Acs et al., 2007; Acs and Armington, 2004). Regarding the other variables, GDP growth and the presence of small firms stimulate regional new firm formation activity as well. Unemployment rate and sunk costs, however, were not found to exert any significant effect.

The key finding of the present research is that of the importance of intra-sectoral knowledge rather than inter-sectoral spillovers. This accords with those studies underlining the industry specific nature of R&D spillovers (Anselin et al., 2000a; 2000b). Our findings on the effect of related and unrelated variety is heavily conditioned on the higher than usual level of aggregation used which was dictated by data availability. It is hoped that future research on regional entrepreneurship in Greece will be able to revisit some of the issues by taking advantage of more disaggregated reliable data, both spatially and sectorally. The results obtained, however, can be reconciled with earlier Greek evidence and it is comparable to evidence obtained in other country contexts.

If specialization is more conducive to regional new formation, then future research needs to examine this in relation to regional structural change and explore its consequences for regional restructuring. This is important if regional entrepreneurship is thought to be a vehicle of regional change and restructuring.

Efforts to increase regional entrepreneurship rates may benefit from a concentration on policy efforts to enhance knowledge-spillover variables. In this direction, one could see efforts facilitating research in universities and encouraging R&D in the private sector as well as by the means of facilitating the collaboration between universities and research institutes on the one hand, and business firms on the other, at the regional level, something in which Greece seems to be lagging behind.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Aarstad

Kvitastein

Jakobsen

(2016) Related and unrelated variety as regional drivers of enterprise productivity and innovation: A multilevel study. Research Policy 45(4): 844–856.

Abel

Deitz

(2012) Do colleges and universities increase their region's human capital? Journal of Economic Geography 12(3): 667–691.

Acs

Armington

(2002) The determinants of regional variation in new firm formation. Regional Studies 36(1): 33–45.

Acs

Armington

(2004) The impact of geographic differences in human capital on service firm formation rates. Journal of Urban Economics 56(2): 244–278.

Acs

Varga

(2005) Entrepreneurship, agglomeration and technological change. Small Business Economics 24(3): 323–334.

Acs

Armington

Zhang

(2007) The determinants of new-firm survival across regional economies: The role of human capital stock and knowledge spillover. Papers in Regional Science 86(3): 367–391.

Acs

Audretsch

Lehmann

(2013) The knowledge spillover theory of entrepreneurship. Small Business Economics 41(4): 757–774.

Acs

Braunerhjelm

Audretsch

et al . (2009) The knowledge spillover theory of entrepreneurship. Small Business Economics 32(1): 15–30.

Agarwal

Audretsch

Sarkar

(2010) Knowledge spillovers and strategic entrepreneurship. Strategic Entrepreneurship Journal 4(4): 271–283.

10.

Alcacer

Chung

(2007) Location strategies and knowledge spillovers. Management Science 53(5): 760–776.

11.

Alexiadis

Tsagdis

(2006) Examining the location factors of R&D labor in the regions of Greece. The Annals of Regional Science 40(1): 43–54.

12.

Anselin

Varga

Acs

(1997) Local geographic spillovers between university research and high technology innovations. Journal of Urban Economics 42(3): 422–448.

13.

Anselin

Varga

Acs

(2000a) Geographical spillovers and university research: A spatial econometric perspective. Growth and Ghange 31(4): 501–515.

14.

Anselin

Varga

Acs

(2000b) Geographic and sectoral characteristics of academic knowledge externalities. Papers in Regional Science 79(4): 435–443.

15.

Anyadike-Danes

Hart

(2006) The impact of sector, specialization, and space on business birth rates in the United Kingdom: A challenge for policy? Environment and Planning C: Politics and Space 24(6): 815–826.

16.

Arrow

(1962) The economic implications of learning by doing. Review of Economic Studies 29(3): 155–173.

17.

Audretsch

Feldman

(1996) R&D spillovers and the geography of innovation and production. American Economic Review 86(3): 630–640.

18.

Audretsch

Fritsch

(1994) The geography of firm births in Germany. Regional Studies 28(4): 359–365.

19.

Audretsch

Keilbach

(2007) The theory of knowledge spillover entrepreneurship. Journal of Management Studies 44(7): 1242–1254.

20.

Audretsch

Lehmann

(2005) Does the knowledge spillover theory of entrepreneurship hold for regions? Research Policy 34(8): 1191–1202.

21.

Audretsch

Stephan

(1996) Company-scientist locational links: The case of biotechnology. American Economic Review 86(3): 641–652.

22.

Audretsch

Stephan

(1999) Knowledge spillovers in biotechnology: Sources and incentives. Journal of Evolutionary Economics 9(1): 97–107.

23.

Audretsch

Lehmann

Warning

(2005) University spillovers and new firm location. Research Policy 34(7): 1113–1122.

24.

Audretsch

Dohse

Niebuhr

(2010) Cultural diversity and entrepreneurship: A regional analysis for Germany. The Annals of Regional Science 45(1): 55–85.

25.

Baltzopoulos

Braunerhjelm

Tikoudis

(2016) Spin-offs: Why geography matters. Journal of Economic Geography 16(2): 273–303.

26.

Bania

Eberts

Fogarty

(1993) Universities and the startup of new companies: Can we generalize from Route 128 and Silicon Valley? Review of Economics and Statistics 75(4): 761–765.

27.

Beaudry

Schiffauerova

(2009) Who's right, Marshall or Jacobs? The localization versus urbanization debate. Research Policy 38(2): 318–337.

28.

Beck

(2001) Time series cross-section data: What have we learned in the last few years? Review of Political Science 4(1): 271–293.

29.

Bishop

(2012) Knowledge, diversity and entrepreneurship: A spatial analysis of subregions in Great Britain. Entrepreneurship and Regional Development 24(7–8): 641–660.

30.

Bishop

Brand

(2014) Human capital, diversity, and new firm formation. The Service Industries Journal 34(7): 567–583.

31.

Boschma

Minondo

Navarro

(2013) The emergence of new industries at the regional level in Spain: A proximity approach based on product relatedness. Economic Geography 89(1): 29–51.

32.

Bosma

Van Stel

Suddle

(2008) The geography of new firm formation: Evidence from independent start-ups and new subsidiaries in the Netherlands. International Entrepreneurship and Management Journal 4(2): 129–146.

33.

Bottazzi

Peri

(2003) Innovation and spillovers in regions: Evidence from European patent data. European Economic Review 47(4): 687–710.

34.

Cainelli

Ganau

Iacobucci

(2016) Do geographic concentration and vertically-related variety foster firm productivity? Micro-evidence from Italy. Growth and Change 47(2): 197–217.

35.

Cameron

Trivedi

(2005) Microeconometrics: Methods and Applications. Cambridge: Cambridge University Press.

36.

Caragliu

De Dominicis

De Groot

HLF

(2016) Both Marshall and Jacobs were right! Economic Geography 92 (1): 87–111.

37.

Carlino

Chatterjee

Hunt

(2007) Urban density and rate of inventions. Journal of Urban Economics 61(3): 389–419.

38.

Carlsson

Acs

Audretsch

et al . (2009) Knowledge creation, entrepreneurship, and economic growth: A historical review. Industrial and Corporate Change 18(6): 1193–1229.

39.

Caves

Porter

(1977) From entry barriers to mobility barriers: Conjectural decisions and contrived deterrence to new competition. The Quarterly Journal of Economics 91(2): 241–262.

40.

Clark

Wrigley

(1997a) The spatial configuration of the firm and the management of sunk costs. Economic Geography 73(3): 285–304.

41.

Clark

Wrigley

(1997b) Exit, the firm and sunk costs: Reconceptualizing the corporate geography of disinvestment and plant closure. Progress in Human Geography 21(3): 338–358.

42.

Clemens

Radelet

Bhavnani

et al . (2012) Counting chickens when they hatch: Timing and the effects of aid on growth. The Economic Journal 122(561): 590–617.

43.

Coll-Martinez

Arauzo-Carod

(2017) Creative milieu and firm location: An empirical appraisal. Environment and Planning A 49(7): 1613–1641.

44.

Corradini

De Propris

(2015) Technological diversification and new innovators in European regions: Evidence from patent data. Environment and Planning A 47(10): 2170–2186.

45.

Daskalopoulou

Liargovas

(2010) Regional determinants of manufacturing start-ups in Greece: Evidence on the effect of agglomeration economies. Applied Economics Letters 17(18): 1841–1844.

46.

Desrochers

Sautet

(2008) Entrepreneurial policy: The case of regional specialization vs. spontaneous industrial diversity. Entrepreneurship Theory and Practice 32 (5): 813–832.

47.

Doring

Schnellenbach

(2006) What do we know about geographical knowledge spillovers and regional growth?: A survey of the literature. Regional Studies 40(3): 375–395.

48.

Eaton

Lipsey

(1980) Exit barriers are entry barriers: The durability of capital as a barrier to entry. Bell Journal of Economics 11(2): 721–729.

49.

Eaton

Lipsey

(1981) Capital, commitment and entry equilibrium. Bell Journal of Economics 12(2): 593–604.

50.

Eurostat (2017) R&D expenditure in the EU remained stable in 2016 at just over 2% of GDP. Eurostat Newsrelease 183/217.

51.

Feldman

Audretsch

(1999) Innovation in cities: Science-based diversity, specialization and localized competition. European Economic Review 43(2): 409–429.

52.

Fotopoulos

(2014) On the spatial stickiness of UK new firm formation rates. Journal of Economic Geography 14(3): 651–679.

53.

Fotopoulos

Louri

(2000) Location and survival of new entry. Small Business Economics 14(4): 311–321.

54.

Fotopoulos

Spence

(1999) Spatial variations in new manufacturing plant openings: Some empirical evidence from Greece. Regional Studies 33(3): 219–229.

55.

Frenken

Van Oort

Verburg

(2007) Related variety, unrelated variety and regional economic growth. Regional Studies 41 (5): 685–697.

56.

Fritsch

Franke

(2004) Innovation, regional knowledge spillovers and R&D cooperation. Research Policy 33(2): 245–255.

57.

Ganau

(2016) Productivity, credit constraints and the role of short-run localization economies: Micro-evidence from Italy. Regional Studies 50(11): 1834–1848.

58.

Garofoli

(1994) New firm formation and regional development: The Italian case. Regional Studies 28(4): 381–393.

59.

Ghio

Guerini

Lehmann

et al . (2015) The emergence of the knowledge spillover theory of entrepreneurship. Small Business Economics 44(1): 1–18.

60.

Glaeser

Kallal

Scheinkman

et al . (1992) Growth in cities. Journal of Political Economy 100(6): 1127–1152.

61.

Goncalves

Almeida

(2009) Innovation and spatial knowledge spillovers: Evidence from Brazilian patent data. Regional Studies 43(4): 513–528.

62.

Gould

Keeble

(1984) New firms and rural industrialization in East Anglia. Regional Studies 18(3): 189–201.

63.

Green

Malpezzi

Mayfo

(2005) Metropolitan-specific estimates of the Price elasticity of supply of housing, and their sources. American Economic Review 95(2): 334–339.

64.

Grillitsch

Nilsson

(2017) Firm performance in the periphery: On the relation between firm-internal knowledge and local knowledge spillovers. Regional Studies 51(8): 1219–1231.

65.

Henderson

Kuncoro

Turner

(1995) Industrial development in cities. Journal of Political Economy 103(5): 1067–1090.

66.

Hoover

(1941) Interstate redistribution of population, 1850–1940. Journal of Economic History 1(2): 199–205.

67.

Jacobs

(1969) The Economy of Cities. New York: Vintage.

68.

Jaffe

(1986) Technological opportunity and spillovers of R&D: Evidence from firms' patents, profits, and market value. American Economic Review 76(5): 984–1001.

69.

Jaffe

(1989) Real effects of academic research. American Economic Review 79(5): 957–970.

70.

Jaffe

Trajtenberg

Henderson

(1993) Geographic localization of knowledge spillovers as evidenced by patent citations. The Quarterly Journal of Economics 108(3): 577–598.

71.

Johansson

Karlsson

(2007) R&D accessibility and regional export diversity. The Annals of Regional Science 41(3): 501–523.

72.

Keeble

Walker

(1994) New firms, small firms and dead firms: Spatial patterns and determinants in the United Kingdom. Regional Studies 28(4): 411–427.

73.

Kirchhoff

Newbert

Hasan

et al . (2007) The influence of university R&D expenditures on new business formations and employment growth. Entrepreneurship Theory and Practice 31(4): 543–559.

74.

Knack

(1993) The voter participation effects of selecting jurors from registration lists. Journal of Law and Economics 36(1): 99–114.

75.

Knoben

Ponds

Van Oort

(2011) Employment from new firm formation in the Netherlands: Agglomeration economies and the knowledge spillover theory of entrepreneurship. Entrepreneurship and Regional Development 23(3–4): 135–157.

76.

Körösi

Mátyás

Székely

(1992) Practical Econometrics. Avebury: Aldershot.

77.

Krugman

(1991) Geography and Trade. Cambridge: MIT Press.

78.

Kutner

Nachtsheim

Neter

(2004) Applied Linear Regression Models. New York: Mc Graw-Hill Irwin.

79.

Lasch

Robert

Le Roy

(2013) Regional determinants of ICT new firm formation. Small Business Economics 40(3): 671–686.

80.

Lee

Acs

Florida

(2004) Creativity and entrepreneurship: A regional analysis of new firm formation. Regional Studies 38(8): 879–891.

81.

Marshall

(1920) Principles of Economics. London: Macmillan.

82.

Maskell

Malmberg

(1999) The competitiveness of firms and regions: Ubiquitification and the importance of localized learning. European Urban and Regional Studies 6(1): 9–25.

83.

Moreno

Paci

Usai

(2005) Spatial spillovers and innovation activity in European regions. Environment and Planning A 37(10): 1793–1812.

84.

Oaxaca

R L

Geisler

(2003) Fixed effects models with time invariant variables: A theoretical note. Economics Letters 80(3): 373–377.

85.

Ouwersloot

Rietveld

(2000) The geography of R&D: Tobit analysis and a Bayesian approach to mapping R&D activities in the Netherlands. Environment and Planning A 32(9): 1673–1688.

86.

Qian

Acs

Stough

(2013) Regional systems of entrepreneurship: The nexus of human capital, knowledge and new firm formation. Journal of Economic Geography 13(4): 559–587.

87.

Reynolds

Storey

Westhead

(1994) Cross-national comparisons of the variation in new firm formation rates. Regional Studies 28(4): 443–456.

88.

Reynolds

Miller

Maki

(1995) Explaining regional variation in business births and deaths: U.S. 1976–88. Small Business Economics 7(5): 389–407.

89.

Rodríguez-Pose

Crescenzi

(2008) R&D spillovers, innovation systems and the genesis of regional growth in Europe. Regional Studies 42(1): 51–67.

90.

Rodríguez-Pose

Hardy

(2015) Cultural diversity and entrepreneurship in England and Wales. Environment and Planning A 47(2): 392–411.

91.

Rogerson

(2001) Statistical Methods for Geography. London: Sage.

92.

Romer

(1986) Increasing returns and long-run growth. Journal of Political Economy 94(5): 1002–1037.

93.

Romer

(1990) Endogenous technology change. Journal of Political Economy 98(5): 71–102.

94.

Roper

Love

Bonner

(2017) Firms’ knowledge search and local knowledge externalities in innovation performance. Research Policy 46(1): 43–56.

95.

Serrano

Cabrer

(2004) The effect of knowledge spillovers on productivity growth inequalities in Spanish regions. Environment and Planning A 36(4): 731–753.

96.

Sonn

Storper

(2008) The increasing importance of geographical proximity in knowledge production: An analysis of US patent citations, 1975–1997. Environment and Planning A: Economy and Space 40(5): 1020–1039.

97.

Stam

(2007) Why butterflies don’t leave: Locational behavior of entrepreneurial firms. Economic Geography 83(1): 27–50.

98.

Theil

(1972) Statistical Decomposition Analysis: With Applications in the Social and Administrative Sciences. Amsterdam: North-Holland.

99.

Van Der Panne

(2004) Agglomeration externalities: Marshall versus Jacobs. Journal of Evolutionary Economics 14(5): 593–604.

100.

Van Stel

Nieuwenhuijsen

(2004) Knowledge spillovers and economic growth: An analysis using data of Dutch regions in the period 1987–1995. Regional Studies 38(4): 393–407.

101.

Varga

Schalk

(2004) Knowledge spillovers, agglomeration and macroeconomic growth: An empirical approach. Regional Studies 38(8): 977–989.

102.

Vogiatzoglou

Tsekeris

(2013) Spatial agglomeration of manufacturing in Greece: Sectoral patterns and determinants. European Planning Studies 21(12): 1853–1872.

103.

Woodward

Figueiredo

Guimaraes

(2006) Beyond the Silicon Valley: University R&D and high-technology location. Journal of Urban Economics 60(1): 15–32.

104.

Zucker

Darby

Brewer

(1998) Intellectual human capital and the birth of U.S. biotechnology enterprises. American Economic Review 88(1): 290–306.